{
  "id": 607329,
  "title": "Goal of competition",
  "url": "/competitions/brain-to-text-25/discussion/607329",
  "author_name": "",
  "post_date": "2025-09-13T10:54:48.771480300Z",
  "votes": null,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Not a major issue, but I noticed in the article you reported a 97.5% accuracy, while in the 2024 competition the best result was a 5.77% error rate.</p>\n<p>For this competition, should we aim to surpass the article’s reported accuracy or the benchmark from the prior competition? I guess it’s the former, but I just wanted to clarify.</p>",
  "messages": [
    {
      "id": "3288199",
      "postDate": "09/13/2025 10:54:48",
      "content": "<p>Not a major issue, but I noticed in the article you reported a 97.5% accuracy, while in the 2024 competition the best result was a 5.77% error rate.</p>\n<p>For this competition, should we aim to surpass the article’s reported accuracy or the benchmark from the prior competition? I guess it’s the former, but I just wanted to clarify.</p>",
      "rawMarkdown": "Not a major issue, but I noticed in the article you reported a 97.5% accuracy, while in the 2024 competition the best result was a 5.77% error rate.\n\nFor this competition, should we aim to surpass the article’s reported accuracy or the benchmark from the prior competition? I guess it’s the former, but I just wanted to clarify.",
      "votes": null
    },
    {
      "id": "3288330",
      "postDate": "09/13/2025 17:40:28",
      "content": "<p>The data split for this competition is unique and not the same as the online performance numbers reported in the NEJM article. The \"UCD-NPL causal RNN + 5gram\" benchmark on this competition's leaderboard uses (largely) the same decoding algorithm as was used in the article.</p>\n<p>For this competition, your goal is to beat the benchmarks and the other teams' submissions on the leaderboard.</p>",
      "rawMarkdown": "The data split for this competition is unique and not the same as the online performance numbers reported in the NEJM article. The \"UCD-NPL causal RNN + 5gram\" benchmark on this competition's leaderboard uses (largely) the same decoding algorithm as was used in the article.\n\nFor this competition, your goal is to beat the benchmarks and the other teams' submissions on the leaderboard.",
      "votes": null
    },
    {
      "id": "3288558",
      "postDate": "09/14/2025 09:15:25",
      "content": "<p>That clarifies it, thank you!</p>",
      "rawMarkdown": "That clarifies it, thank you!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3288330,
      "author_name": "notnickc",
      "author_url": "",
      "post_date": "09/13/2025 17:40:28",
      "content": "<p>The data split for this competition is unique and not the same as the online performance numbers reported in the NEJM article. The \"UCD-NPL causal RNN + 5gram\" benchmark on this competition's leaderboard uses (largely) the same decoding algorithm as was used in the article.</p>\n<p>For this competition, your goal is to beat the benchmarks and the other teams' submissions on the leaderboard.</p>",
      "votes": null,
      "replies": [
        {
          "id": 3288558,
          "author_name": "elhnan",
          "author_url": "",
          "post_date": "09/14/2025 09:15:25",
          "content": "<p>That clarifies it, thank you!</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "3288199": "Not a major issue, but I noticed in the article you reported a 97.5% accuracy, while in the 2024 competition the best result was a 5.77% error rate.\n\nFor this competition, should we aim to surpass the article’s reported accuracy or the benchmark from the prior competition? I guess it’s the former, but I just wanted to clarify.",
    "3288330": "The data split for this competition is unique and not the same as the online performance numbers reported in the NEJM article. The \"UCD-NPL causal RNN + 5gram\" benchmark on this competition's leaderboard uses (largely) the same decoding algorithm as was used in the article.\n\nFor this competition, your goal is to beat the benchmarks and the other teams' submissions on the leaderboard.",
    "3288558": "That clarifies it, thank you!"
  },
  "source": "meta"
}